Also this week: how to become really highly cited, robot statistician, universities vs. brands, and more.
We’re a bit late to this: the 100 most cited scientific papers ever. See how many you can guess before clicking through. Hint: They’re mostly methods papers. Hint #2: There are no ecology papers (not even close!). A few papers on statistical methods and software packages for phylogenetic estimation are the only evolution papers that make the cut. Of course, really influential work is rarely cited, as it’s just part of the background knowledge every scientist is supposed to have. If that weren’t the case, R. A. Fisher would probably be the most-cited scientist ever.
Speaking of citations: according to this preprint, a (modestly) greater fraction of citations now go to papers >10 (or 15, or 20) years old than was the case in 1990. Most areas of scholarship show the trend, albeit to varying degrees. The study’s based on Google Scholar data; I’m not sure if that creates any artifacts. (ht Marginal Revolution)
How to spot the holes in a data-based news story. Very good reading for undergrad stats classes. Based on compelling real world examples. Particularly good on driving home the points that correlation is not causation, and the reasons why statistically controlling for confounding variables often is ineffective.
Good tips for giving a good talk. Includes some advice I haven’t seen elsewhere.
Jeremy Yoder has an interview with the founders of Haldane’s Sieve, a website that promotes and discusses preprints in population and evolutionary genetics. Always interesting to hear from folks who are experimenting with new ways of doing things. Glad to see that they recognize a key virtue of the current pre-publication peer review system: it ensures that at least some close attention is paid to every paper. I’m pessimistic that there’s any way to prevent serious attention concentration post-publication (see also here). Indeed, isn’t Haldane’s Sieve itself a mechanism for concentrating post-publication attention? I was also interested in their perception that the scientific publication system is mostly an overly-critical, “down-voting” system. That might be true of pre-publication review, but if anything I think the opposite is the case post-publication. Post-publication, bandwagons and zombie ideas far outnumber Buddy Holly ideas, and even clear-cut mistakes continue to attract attention and citations much longer than they should. So if you want a scientific publication system that achieves some ideal balance of “up voting” enthusiasm and “down voting” criticism, well, maybe our current system isn’t too far off the mark? Our current pre-publication review system also has the advantage that it is a system, with agreed, enforced rules and norms that at least in principle (and I think for the most part in practice) apply equally to everybody, and that everybody knows they’re signing up for when they start doing science (see here for discussion). The next person who figures out what the rules and norms of post-publication commenting should be (in particular, what the rules and norms for critical comments should be), gets everyone to agree to them, and figures out how to enforce them, will be the first.
A while back I discussed the suggestion for a “deterministic statistical machine”–basically, statistical software that would automatically choose an appropriate analysis and then do it for you. It would be aimed at users who don’t know statistics, much as premade meals are aimed at people who can’t (or won’t) cook. Now someone’s invented such a machine.
What to do if you’ve been denied tenure, or are about to be. Related: Meg’s old post on how to navigate the tenure track and maximize the odds that you never need to click that link (although once you have a tenure-track faculty position, the odds are very much in your favor).